Friday, January 30th 2026

AMD "Zen 6" CCD Packs 12 Cores, 48 MB L3 Cache

The standard CPU complex die (CCD) chiplet based on AMD's next-generation "Zen 6" microarchitecture is poised to feature a 50% increase in CPU core counts. This would be the first time that AMD has increased core-counts of its CCDs containing full-sized CPU cores capable of sustaining high clock speeds. It had developed CCDs with high core counts using compacted cores with lower clock speed limits until now. The standard "Zen 6" CCD is expected to come with a similar die-size to the current "Zen 5" CCD, with a rumored die size of 76 mm², which is slightly larger than the 71 mm² of the "Zen 5" CCD.

The "Zen 6" CCD is expected to pack 12 CPU cores, all part of a single CPU core complex (CCX). All twelve cores share a 48 MB L3 cache. Both the CPU core counts and L3 cache size hence see a 50% increase over "Zen 5." AMD is building the "Zen 6" CCD on TSMC N2 (2 nm nanosheet) process node, which evidently offers a massive increase in transistor densities over the current TSMC N4P (4 nm FinFET) on which the "Zen 5" CCD is made. With an eye on Intel's pivot towards a cache rebalancing in favor of large last-level caches on some of its premium Core Ultra 400 "Nova Lake-S" desktop processors, AMD is expected to give the "Zen 6" CCD full 3D V-Cache capability, and the company's next X3D processors could come with L3 caches of 144 MB per CCD for up to 288 MB of L3 cache on the desktop AM5 socket.
Sources: HXL (Twitter), VideoCardz
Add your own comment

99 Comments on AMD "Zen 6" CCD Packs 12 Cores, 48 MB L3 Cache

#26
shuffy_losys
I find it amusing that AMD has been allowed to basically retain the same core count setup for almost a decade and the only reason it was able to do so was because it kept up enough of a per-core performance uplift for it to matter.
Posted on Reply
#27
R0H1T
Nah, it was because of price ~ here at its cheapest, last year, I could buy a 5950x for under $300 including EU like (high) taxes! Find me a better value especially given the cost of "AM4" as a platform.
Posted on Reply
#28
ymdhis
Mr. PerfectThe real question is how long are they going to make everyone wait for the X3D parts? They've traditionally been a year or so after the non-X3D parts as some sort of product segmentation play, which is extremely annoying for gamers who aren't even considering non-X3D parts.
Oh you think waiting for X3D parts is bad, try waiting for their APU releases. Half-assed mobile chip ports every 2-3 years, and as of Phoenix they also have half the PCIE lanes and non-soldered IHS.
Posted on Reply
#29
shuffy_losys
R0H1TNah, it was because of price ~ here at its cheapest, last year, I could buy a 5950x for under $300 including EU like (high) taxes! Find me a better value especially given the cost of "AM4" as a platform.
I don't think value proposition matters on a part that's that old. Especially when you consider that Zen 3 was the first generation AMD bumped up the price because they said they were the performance leaders back then and they could justify it.

EDIT: Out of curiosity, I wanted to see how that could compare with a more modern $300 process. And the 9600X appears to hold its own pretty well against the 5950X.

Sure the cost of the rest of the system is/was probably cheaper but isolating the CPUs, it sort of drives my point that AMD could get away with not bumping up core counts because per-core performance makes up for it.
Posted on Reply
#30
ARF
shuffy_losysEDIT: Out of curiosity, I wanted to see how that could compare with a more modern $300 process. And the 9600X appears to hold its own pretty well against the 5950X.

Sure the cost of the rest of the system is/was probably cheaper but isolating the CPUs, it sort of drives my point that AMD could get away with not bumping up core counts because per-core performance makes up for it.
The poor multi-thread app optimisation hides the otherwise relatively significant performance advantage of the 32-thread Ryzen 9 5950X against the 12-thread Ryzen 5 9600X.
In properly optimised software the former is around 50% faster than the latter.



Posted on Reply
#31
MufasaV
R0H1TIt isn't just that, doubling cache doesn't guarantee anything although in practice you would get slightly ~ massive performance increase in most applications. Otherwise Intel, Nvidia, QC, Apple would simply be doubling their L1//L2/L3 caches annually & that'd be it! It's usually a tradeoff between (die) space, latency, clocks, power consumption et al.
Nvidia more than 10X their L2 cache in one Gen. That was after AMD did the same with L3 cache. GPUs. CPUs are the sane issue for AMD X3D. Intel are the only ones who haven't tied it in any product and this stand the most to gain from it despite trade offs
Posted on Reply
#32
Rulumi
If true, nice and I think it is time for at least an small core increment, it's already going to be 9 years of AMD keeping the same core count in most of their desktop consumer CPUs, it's going to become longer than the time Intel took with the dual and quad cores segmentation in the main consumer platforms. And honestly at least in the low end and mid segments it doesn't feel like they are keeping up that well anymore, while low end can make sense if the price is fine and for people that don't need much multi-core performance or plan to upgrade later on.

Right now the current value of the Ryzen 7 like the 7700/X and the 9700/X feels quite low when the Core Ultra 5 235 and 245K are around the same performance in most things while costing less and the gap with the Core Ultra 7 265/K in most things just grows bigger and right now it has a very good price. And this isn't that new either, it was and is already the same case against the Core i5 14600K and the Core i7 14700/K. If Nova Lake really is bringing back AVX-512 and technically even a bit better because the AVX10.2, which is an advantage that helps AMD performance in some applications right now, alongside some other said things like here and other reports, Nova Lake and Zen 6 could be an oppurtinity of interesting competition for consumers.
RedwoodzThat all sounds right untill you realise the Intel chip takes 280mm sq. to outperform an 84mm sq. chip.
It costs less, is more power efficient, LGA 1851 motherboards have good prices right now and part of the space is taken by the much better and useful iGPU on those than any of the current AMD integrated graphics options in desktop.
Posted on Reply
#33
Dr_b_
can we please have more PCIe Lanes? Ill trade 4 cores for another 16 PCIe 5.0 lanes
Posted on Reply
#34
zenlessyank
Stick 128 Gigs HBM on there and I will be happy.
Posted on Reply
#35
shuffy_losys
ARFThe poor multi-thread app optimisation hides the otherwise relatively significant performance advantage of the 32-thread Ryzen 9 5950X against the 12-thread Ryzen 5 9600X.
In properly optimised software the former is around 50% faster than the latter.
You act as though all apps can be "properly" multithreaded the same as an app that runs an embarrassingly parallel task. But sure, I'll bite that in such scenarios, more cores is better, though not necessarily threads considering an i7-13700K can beat a 5950X in Cinebench 24 despite having fewer threads.
RedwoodzThat all sounds right untill you realise the Intel chip takes 280mm sq. to outperform an 84mm sq. chip.
ARL's compute die size was estimated to be around 114.5mm^2. And that includes the 16 E-cores.
Dr_b_can we please have more PCIe Lanes? Ill trade 4 cores for another 16 PCIe 5.0 lanes
What are you going to use it for? Multi-GPU is effectively dead and unless you're running a database center, storage drives are going to be idling 99% of the time. And it's not like the bandwidth actually matters in most other applications.
Posted on Reply
#36
ARF
shuffy_losysYou act as though all apps can be "properly" multithreaded the same as an app that runs an embarrassingly parallel task. But sure, I'll bite that in such scenarios, more cores is better, though not necessarily threads considering an i7-13700K can beat a 5950X in Cinebench 24 despite having fewer threads.
I remember how they changed the base CPU-Z benchmark right after Ryzen was launched back in Q2 2017, because Intel was trailing, and Intel needed a benchmark that suits their agenda first. We know some software just detect Intel and magically run better.
More cores is better, especially if you run several apps side by side. How many apps do you think a hexa-core CPU can run? 2 apps, 3 apps? And now compare with how many apps a 16-core CPU can run simultaneously. This means gaming, streaming, coding, rendering, all at the same time, side by side.
Posted on Reply
#37
Dr_b_
shuffy_losysWhat are you going to use it for? Multi-GPU is effectively dead and unless you're running a database center, storage drives are going to be idling 99% of the time. And it's not like the bandwidth actually matters in most other applications.
from a NIC, to a sound card, to a nvme drives, to mcio connectors, to not having to have 2 chipsets to extend anemic IO on motherboards to allow for more USB ports/TB ports

What would i use it for? adding a multiport high speed NIC, a sound card, PCIe to nvme card, or m.2 nvme slots on the motherboard that dont steal bandwidth from the only 16X slot on motherboards these days, and anything else in a PCIe form factor, and yes, potentially even another GPU. Multi GPU isn't dead, you can scale workloads to multiple GPUs just fine.
Posted on Reply
#38
Unregistered
shuffy_losysWhat are you going to use it for? Multi-GPU is effectively dead and unless you're running a database center, storage drives are going to be idling 99% of the time. And it's not like the bandwidth actually matters in most other applications.
More NVME, maybe a PCIE capture card, soundcard, etc
Posted on Edit | Reply
#39
shuffy_losys
Dr_b_from a NIC, to a sound card, to a nvme drives, to mcio connectors, to not having to have 2 chipsets to extend anemic IO on motherboards to allow for more USB ports/TB ports

What would i use it for? adding a multiport high speed NIC, a sound card, PCIe to nvme card, or m.2 nvme slots on the motherboard that dont steal bandwidth from the only 16X slot on motherboards these days, and anything else in a PCIe form factor, and yes, potentially even another GPU. Multi GPU isn't dead, you can scale workloads to multiple GPUs just fine.
ShadowizedMore NVME, maybe a PCIE capture card, soundcard, etc
I would put down some counter arguments but ultimately the only counter argument that holds any weight here is most consumers almost never expand what's in their computers, assuming they even get a desktop. I've seen more computer with empty slots than I've seen stuffed to the gills. I'm not saying they don't exist, but they're the outliers, and manufacturers only care about the majority. And if the majority isn't going to use more than what they offer now, then the manufacturers have no real incentive to provide more. It's likely the reason why HEDT platforms died off: only a handful of diehards actually cared and it wasn't enough to sustain the product line.

In any case, I could also be pedantic and say that since this news article is about the CCDs, it doesn't matter anyway. They don't have PCIe root complexes.
Posted on Reply
#40
Unregistered
shuffy_losysI would put down some counter arguments but ultimately the only counter argument that holds any weight here is most consumers almost never expand what's in their computers, assuming they even get a desktop. I've seen more computer with empty slots than I've seen stuffed to the gills. I'm not saying they don't exist, but they're the outliers, and manufacturers only care about the majority. And if the majority isn't going to use more than what they offer now, then the manufacturers have no real incentive to provide more. It's likely the reason why HEDT platforms died off: only a handful of diehards actually cared and it wasn't enough to sustain the product line.

In any case, I could also be pedantic and say that since this news article is about the CCDs, it doesn't matter anyway. They don't have PCIe root complexes.
I feel your point is only valid for the lowend chipsets ie b650 b850 etc, which are what the consumers who don't upgrade anything are going to be using. It should not be a thing on X870E and beyond. A couple of extra lanes to give a soundcard/capture card or another NVME some headroom is hardly asking for the world.
Posted on Edit | Reply
#42
Launcestonian
Zen 6 going to give us upgraded memory controller?.. I want 8000MT/S 1:1 :)
Posted on Reply
#43
LastDudeALive
Dr_b_from a NIC, to a sound card, to a nvme drives, to mcio connectors, to not having to have 2 chipsets to extend anemic IO on motherboards to allow for more USB ports/TB ports

What would i use it for? adding a multiport high speed NIC, a sound card, PCIe to nvme card, or m.2 nvme slots on the motherboard that dont steal bandwidth from the only 16X slot on motherboards these days, and anything else in a PCIe form factor, and yes, potentially even another GPU. Multi GPU isn't dead, you can scale workloads to multiple GPUs just fine.
Unfortunately, every single thing you mentioned is extremely niche. Most users don't even use a NIC or are unaware they can. WiFi is internet, to them. Sound cards are the same, 99% of people plug their headphones into the motherboard, monitor, or use Bluetooth. 99% of users have no use for more than 2 or three NVME drives, if they're using multiple at all. I'm fairly tech-savvy, and I had to Google what MCIO is, I've literally never heard of it. 99% of users have no need for more USB than what comes on the motherboard. Keyboard, mouse, maybe a webcam, and that's it. Multi-GPU has no use for 99% of the market.

If you want premium features, you pay for the premium Threadripper platform. That means 99% of people aren't paying extra to subsidize the extreme minority of power users.
Posted on Reply
#44
_roman_
not for the corresponding am5 zen6 apu for sure. the leaker is _roman_
Posted on Reply
#45
FoulOnWhite
LastDudeALiveUnfortunately, every single thing you mentioned is extremely niche. Most users don't even use a NIC or are unaware they can. WiFi is internet, to them. Sound cards are the same, 99% of people plug their headphones into the motherboard, monitor, or use Bluetooth. 99% of users have no use for more than 2 or three NVME drives, if they're using multiple at all. I'm fairly tech-savvy, and I had to Google what MCIO is, I've literally never heard of it. 99% of users have no need for more USB than what comes on the motherboard. Keyboard, mouse, maybe a webcam, and that's it. Multi-GPU has no use for 99% of the market.

If you want premium features, you pay for the premium Threadripper platform. That means 99% of people aren't paying extra to subsidize the extreme minority of power users.
I have 3 M.2 drives. I had to use specific slots. If i had used others i would have had only x8 on my GPU, or disabled other things on the board. I guess a CPU with more PCIe lanes would be better but much more costly. They put 5 M.2 slots on a board for you not to use all of them?
Posted on Reply
#46
Jtuck9
hsewIt's possible Zen 6 or 7 transitions away from a Performance/Compact core design to just Compact cores, but without the distinction (no more C core designation if they're all the same core again), making scheduling a bit easier. This might negate any performance gains for a generation, but it is the easiest path to shrinking enough space to allow 50% more cores in less than 10% more die space.
Maybe with the updated architecture / controller / an updated scheduler you'll be able to do more with your threads (time slicing)?! Perhaps utilizing your "less" performant cores. Frequency vs bandwidth...
Posted on Reply
#47
AusWolf
Give it a better IO die with seamless 6000-6400 MHz RAM speed support on high capacities, without the need for excessive VSoC, and I'll be happy. This will probably be my last upgrade for the next decade.
Posted on Reply
#48
N3utro
SemperPC09I am very excited for Zen6, it seems like it will be quite good. Also, hopefully HammerLake and all the other -lakes from Intel are competitive so that AMD doesn't slack off. A 12 core for 300$ with a better ipc than zen5 would be perfect. Even a 10 core for 300$ would be banger. I don't need more than about 8 cores, especially if they are relatively modern/have decent ipc.
The recently released 8 core 9850X3D is sold $500, and you think they're going to sell $300 a 12 cores CPU with improved 2nm manufacturing process and cache size?

You're dreaming. More logically it would be sold double that at least.
Posted on Reply
#49
ARF
N3utroThe recently released 8 core 9850X3D is sold $500, and you think they're going to sell $300 a 12 cores CPU with improved 2nm manufacturing process and cache size?

You're dreaming. More logically it would be sold double that at least.
Greed will lead you nowhere - especially when AMD is about to become the underdog once again after the release of the new 52-core Intel CPUs. AMD will have no answer, except lower pricings.
AusWolfGive it a better IO die with seamless 6000-6400 MHz RAM speed support on high capacities, without the need for excessive VSoC, and I'll be happy. This will probably be my last upgrade for the next decade.
Is 6000-6400 too low for DDR5? It must be DDR5-8000, at least.

I won't upgrade with any AMD CPU in the distant future. I have my 24-thread Ryzen 9 5900X and it is very fast in its rated TDP of just 105W.
Posted on Reply
#50
AusWolf
ARFIs 6000-6400 too low for DDR5? It must be DDR5-8000, at least.
I'm not looking for blazing RAM speeds. I'm even fine with 4800 with my 7950X3D. I just want my kit to work at its expo rated speed without too much SoC voltage.
ARFI won't upgrade with any AMD CPU in the distant future. I have my 24-thread Ryzen 9 5900X and it is very fast in its rated TDP of just 105W.
With an AM4 system, I wouldn't upgrade either, as it's still plenty fast. I'm only thinking about upgrading because Zen 6 will be just a CPU drop-in for me. If I needed a new motherboard and RAM, it would be a hard pass.
Posted on Reply
Add your own comment
Nov 17th, 2026 13:51 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts